A new study from researchers at King’s College London and Germany’s Protestant University of Applied Sciences proposes a framework for how AI chatbots might reinforce delusional thinking in vulnerable people.
They call it an “amplification spiral.” The idea: three common chatbot behaviors — linguistic alignment (mirroring your communication style), hyper-personalized responses, and sycophancy (agreeing with you) — combine into a feedback loop that can elaborate and reinforce delusions over time.
The researchers compare it to an “echo chamber of one,” where the normal corrective influence of real-life social interactions is absent. Unlike radio or TV, chatbots engage in prolonged, personalized conversations — which makes the dynamic qualitatively different from past technologies that appeared in delusional beliefs.
The authors are careful to note: no causal link between AI use and psychosis has been established. This is a framework to guide future research, not a proven mechanism.
Still, the timing matters. A recent APA survey found 15% of psychologists reported patients developing distorted thinking related to chatbot use. Lawsuits against OpenAI, Google, and xAI over chatbot-related harms are already working through courts. The question isn’t going away.
